Update app.py

app.py

@@ -50,6 +50,42 @@ max_new_tokens = 2048
 #     for token in model.generate(prompt=prompt, temp=temperature, top_k = top_k, top_p = top_p, max_tokens = max_new_tokens, streaming=True):
 #         outputs.append(token)
 #         yield "".join(outputs)
+
+# async def generater(message, history, temperature, top_p, top_k):
+#     # 构建prompt
+#     prompt = ""
+#     for user_message, assistant_message in history:
+#         prompt += model.config["promptTemplate"].format(user_message)
+#         prompt += assistant_message
+#     prompt += model.config["promptTemplate"].format(message)
+
+#     # Debug: 打印最终的prompt以验证其正确性
+#     print(f"Final prompt: {prompt}")
+#     cmd = [
+#         main_path,
+#         "-m",os.path.join(model_path, model_name),
+#         "--prompt", prompt
+#     ]
+
+#     # 使用subprocess.Popen调用./main并流式读取输出
+#     process = subprocess.Popen(
+#         cmd, stdout=subprocess.PIPE, stderr=subprocess.PIPE, text=True
+#     )
+
+#     # 初始占位符输出
+#     yield "Generating response..."
+#     # 异步等待并处理输出
+#     try:
+#         while True:
+#             line = process.stdout.readline()
+#             if not line:
+#                 break  # 如果没有更多的输出，结束循环
+#             print(f"Generated line: {line.strip()}")  # Debug: 打印生成的每行
+#             yield line
+#     except Exception as e:
+#         print(f"Error during generation: {e}")
+#         yield "Sorry, an error occurred while generating the response."
+
 async def generater(message, history, temperature, top_p, top_k):
     # 构建prompt
     prompt = ""
@@ -59,31 +95,76 @@ async def generater(message, history, temperature, top_p, top_k):
     prompt += model.config["promptTemplate"].format(message)
 
     # Debug: 打印最终的prompt以验证其正确性
-    print(f"Final prompt: {prompt}")
+    print(f"Final prompt: {prompt}\n\n\n\n\n\n\n\n")
     cmd = [
-        main_path,
-        "-m",os.path.join(model_path, model_name),
+        "./main",  # 确保这个是可执行文件的正确路径
+        "-m", os.path.join(model_path, model_name),
         "--prompt", prompt
     ]
 
-    # 使用subprocess.Popen调用./main并流式读取输出
-    process = subprocess.Popen(
-        cmd, stdout=subprocess.PIPE, stderr=subprocess.PIPE, text=True
+    # 创建异步子进程
+    process = await asyncio.create_subprocess_exec(
+        *cmd,
+        stdout=asyncio.subprocess.PIPE,
+        stderr=asyncio.subprocess.PIPE,
+        #text=True,  # 这里设置text=True使得输出以字符串形式处理
     )
 
     # 初始占位符输出
     yield "Generating response..."
-    # 异步等待并处理输出
-    try:
-        while True:
-            line = process.stdout.readline()
-            if not line:
-                break  # 如果没有更多的输出，结束循环
-            print(f"Generated line: {line.strip()}")  # Debug: 打印生成的每行
-            yield line
-    except Exception as e:
-        print(f"Error during generation: {e}")
-        yield "Sorry, an error occurred while generating the response."
+    
+    # # 异步等待并逐字处理输出
+    # while True:
+    #     char = await process.stdout.read(1)  # 读取1字节
+    #     if not char:
+    #         break  # 如果没有更多的输出，结束循环
+    #     # 直接输出字符，这里假设输出是文本形式
+    #     print(char, end='', flush=True)  # 使用print来立即输出每个字符
+    #     yield char
+
+    # while True:
+    #     char = await process.stdout.read(1)  # 读取1字节
+    #     if not char:
+    #         break  # 如果没有更多的输出，结束循环
+    #     # 将字节解码为字符串
+    #     char_decoded = char.decode('utf-8')
+    #     print(char_decoded, end='')  # 使用print来立即输出每个字符
+    #     yield "1"
+
+    # # 等待子进程结束
+    # await process.wait()
+    # 初始化一个空字节串用作缓冲区
+    buffer = b""
+    # 初始化一个空字符串用于累积解码的输出
+    accumulated_output = ""
+
+    while True:
+        # 尝试从stdout中读取更多的字节
+        more_bytes = await process.stdout.read(1)
+        if not more_bytes:
+            break  # 没有更多的字节可以读取，结束循环
+        buffer += more_bytes  # 将新读取的字节添加到缓冲区
+
+        try:
+            # 尝试解码整个缓冲区
+            decoded = buffer.decode('utf-8')
+            # 将成功解码的内容添加到累积的输出中
+            accumulated_output += decoded
+            # 输出累积的内容到屏幕上
+            print(f'\r{accumulated_output}', end='', flush=True)
+            yield accumulated_output
+            buffer = b""  # 清空缓冲区以接受新的输入
+        except UnicodeDecodeError:
+            # 解码失败，可能是因为字节不完整
+            # 继续循环，读取更多的字节
+            continue
+
+    # 循环结束后，处理缓冲区中剩余的字节
+    if buffer:
+        # 这里忽略解码错误，因为最后的字节可能不完整
+        remaining_output = buffer.decode('utf-8', errors='ignore')
+        accumulated_output += remaining_output
+        print(f'\r{accumulated_output}', end='', flush=True)
 
 
 def vote(data: gr.LikeData):